1. Introduction

Single-longitudinal-mode (SLM) ultra-narrow linewidth fiber lasers are the preferred light sources for many cutting-edge applications as per the unique attributes of long coherence length, low noise, simplicity, and compatibility with fiber optic systems. They can be utilized in large-capacity ultra-long-haul coherent optical communication, high-accuracy optical metrology and spectroscopy, high-resolution distributed fiber sensing, coherence Doppler Lidar, laser interferometer gravitational-wave observatory [1–6], and other applications related to optical atomic clocks, measurements of fundamental constants, and physics [7–10]. Multi-functional SLM fiber lasers capable of wavelength-tuning/sweeping/switching, dual-/multi-wavelength operation, Q-switching, and polarization-state transforming have shown remarkable potential for the photonic generation of high spectral purity microwave or terahertz (THz) waves, and may be applicable in multi-parameter distributed fiber optic sensing, free-space optical communication, and wavelength-division-multiplexing systems [11–14]. The short laser cavity configurations, including distributed feedback (DFB) and distributed Bragg reflector (DBR), are the simplest techniques for fabricating SLM fiber lasers [11,15,16]. However, the output of such devices is of low power with high-intensity noise and poor function extensibility. The long ring laser cavity configuration allows different functional devices to be introduced into the lasing system for the purposes of multi-functional output and enhanced performances [17–21].

High frequency stabilization is essential for achieving narrow linewidth fiber lasers. Passive stabilizing techniques such as sound insulation, anti-vibration, and constant temperature control are commonly used in fiber laser systems; the frequency instability is limited to >10⁻⁸, which can only guarantee an SLM laser with a linewidth between several kilohertz (kHz) to tens of kHz. For achieving sub-kHz or even narrower laser linewidth, an effective compressing mechanism is necessary. Many linewidth compression techniques have been reported previously including the slow-light effect for extending photon lifetime [22], self-injection locking [23,24], and saturable absorber (SA) based dynamic Bragg grating [19,25]. The slow-light effect and self-injection locking can reduce the linewidth of DFB and DBR fiber lasers significantly but they are less effective for long ring cavity fiber lasers. A microring reflector can provide high-efficient self-injection, but also introduces high insertion loss in the laser system. The unpumped rare-earth-doped fiber-based SA degrades the optical signal-to-noise ratio and Q-value of fiber lasers; it is also ineffective in SLM lasers with linewidths narrower than kHz order.

The Pound-Drever-Hall (PDH) active frequency stabilization method [26] works via locking the laser frequency to an extremely stable, high-finesse reference cavity under the condition of ultralow temperature fluctuation and active feedback control; PDH can reduce the laser frequency instability to a state-of-the-art level of <10⁻¹⁶ [27] and compress the SLM fiber laser’s linewidth to the millihertz (mHz) level [28,29]. However, the PDH system is overly complex, bulky, and expensive, which limits the transportability of locked ultra-stable fiber lasers and prevents low-cost volume production. Locking the laser frequency to a heterodyne Michelson interferometer with a long fiber delay line can reach a short-term frequency instability of 2 × 10⁻¹⁵ and a linewidth of 200 mHz [30]. However, this technique still depends on a complicated photo-electric auxiliary servo system. Although active frequency locking methods can offer outstanding linewidth and stability for SLM fiber lasers, they are only suitable for cost-insensitive applications rather than more common, cost-sensitive applications. There is demand, to this effect, for compressing the linewidth of fiber lasers to the level of tens of Hz to few hundreds of Hz via a low-cost method at room temperature.

The laser linewidth after backscattering is narrower than the incident light in a Rayleigh scattering structure, which has been validated both in theory and experiment by Zhu et al. [31–33]. Based on that, they proposed a new method to significantly compress the laser linewidth using the distributed Rayleigh backscattering in a long single-mode fiber (SMF) [32–37]. By carefully adjusting a variable optical attenuator (VOA) to balance the scattering intensity from the long SMF and the reflection intensity of a Faraday rotator mirror (FRM) in the fiber laser cavity, a laser linewidth as low as 100-200 Hz was achieved [33,36]. This method is both function- and cost-effective. However, adjusting the VOA to balance between the scattering SMF and FRM is impractical; further, the insertion of tens or even hundreds of meters of SMF lengthens the laser cavity significantly, which severely degrades the longitudinal-mode stability against environmental disturbance and temperature variations. In addition, using an overly lengthy SMF may introduce stimulated Brillouin scattering (SBS) for high-power oscillation in the laser cavity. Though a tapered SMF may replace the common SMF to reduce the SBS threshold and increase the scattering coefficient, a length of >100 m tapered SMF is still necessary [37,38]. A short high scattering fiber (HSF) is desired to provide sufficient distributed feedback for laser linewidth compression while maintaining a short laser cavity for stable, ultra-narrow linewidth fiber lasers.

Traditional scattering enhancement to silica optical fiber via high-concentration doping requires a highly complex manufacturing process and can result in high material loss [39]. In recent years, femtosecond laser fabrication has been broadly used in optical fibers to inscribe numerous functional devices [11,40–42]. Some previous researchers have explored scattering enhancement in SMFs [43,44]. However, there has been no previous study on the effectiveness of femtosecond laser-induced HSF for linewidth compression of SLM fiber lasers.

In this work, we explore using the linewidth compression effect of an HSF with randomly distributed high scattering centers induced by femtosecond laser pulses to reduce the linewidth of fiber lasers. A dual-wavelength-switchable SLM erbium-doped fiber laser (EDFL) is used for demonstration, which is mainly composed of a 3 m long erbium-doped fiber (EDF) as gain medium, a dual-coupler-ring based compound-cavity (DCR-CC) filter for SLM selection, and a high-birefringence fiber Bragg grating (HB-FBG) for defining two lasing wavelengths. The effectiveness of linewidth compression as a function of HSF length is examined. We find that the laser linewidth reduces from >1 kHz to <150 Hz as the HSF length increases. The linewidth reduction effect is significant when the laser operates in single-wavelength and dual-wavelength operation modes. Beyond the linewidth, the proposed fiber laser also performs well in terms of wavelength and power stability, optical signal-to-noise ratio (OSNR), longitudinal-mode stability, frequency noise, and relative-intensity-noise (RIN).

2. Experimental setup, principle, and theory

Figure 1 shows a schematic diagram of the proposed EDFL system, which has a very simple configuration of a ring cavity fabricated from low-cost components. A 3 m long EDF (Fibercore Cor., M12-980-125) is used as the gain medium, which is coiled around the plates of a three-loop polarization controller (TL-PC). A 980 nm laser diode (LD) as the pump source provides energy for EDF excitation through a 980/1550 nm wavelength-division-multiplexer (WDM). A DCR-CC filter composed of four optical couplers (OC1, OC2, OC3, and OC4) is used as an ultra-narrow filter for SLM selection. A three-port optical circulator introduces the HB-FBG as a reflecting two-channel wavelength selector, and also functions as an isolator to guarantee the unidirectional propagation of light in the cavity. A certain length of HSF is inserted between the HB-FBG and circulator port 2 for laser linewidth compression. Another coupler OC5 is used to deliver 30% laser output from the cavity for measurement and to maintain 70% energy inside the cavity for oscillation. The whole laser system is placed in an isolation box made of acrylic board to prevent air vibration or acoustic waves from destabilizing the laser, as shown in the inset.

 

Fig. 1. Dual-wavelength-switchable ultra-narrow linewidth erbium-doped fiber laser (EDFL) system using the HSF. LD: laser diode; WDM: wavelength division multiplexer; EDF: erbium-doped fiber; TL-PC: three-loop polarization controller; OC: optical coupler; DI-PC: drop-in polarization controller; DCR-CC: dual-coupler-ring based compound-cavity; HB-FBG: high-birefringence fiber Bragg grating; HSF: high scattering fiber.

The TL-PC is made with a 3 m long EDF pigtailed by single-mode fibers (SMF-28) at both sides. The HB-FBG, combined with the gain EDF coiled inside the TL-PC, introduces enhanced polarization hole burning to mitigate the strong wavelength competition. The DCR-CC filter selects a single lasing mode among dense longitudinal-modes. The HSF compresses the laser linewidth significantly. The FC/APC connecter behind the HB-FBG prevents unnecessary reflection. Inset is a photo of the proposed EDFL system shielded in an air vibration and acoustic wave isolation box made of acrylic board.

Download Full Size | PDF

We investigated the DCR-CC filter in detail in a previous study [21]. Using a previously proposed simulation method [18], we designed the fabrication parameters directly according to the fiber laser’s structural parameters. The crossing-coupling ratios of the four couplers are all 10% and the lengths of two sub-rings OC1&OC2 and OC3&OC4 are 60 cm and 62 cm, respectively. According to the designed parameters, the DCR-CC filter has the transmission spectrum shown in Fig. 2(a), which was measured using a wavelength-swept laser source (Yenista T100S-HP), a 400 MHz photodetector (PD, Thorlabs PDB470C), and a data acquisition card (DAQ, Measurement Computing Cor. USB-1602HS) [18,21]. Figure 2(b) shows the enlarged spectrum of Fig. 2(a) at ∼1556.3 nm indicative of the filtering properties. The free-spectrum-range (FSR) and full-width at half maximum (FWHM) of the main filtering channel are, respectively, 10.31 GHz and 8.42 MHz. The suppression ratio (SR), defined in the caption of Fig. 2, is as low as 0.482 as given in Fig. 2(b). Generally, an SR below 0.5 indicates effective filtering performance in a DCR-CC filter [21]. The HB-FBG was inscribed in a section of hydrogen-loaded polarization maintaining fiber (PMF, YOFC PM#1550_125-18/250) based on the phase mask method, using a uniform phase mask with a period of $\Lambda $=1075 nm and a KrF excimer laser emitting ultraviolet light at 248 nm. The PMF has birefringence $\Delta n$ of ∼2.5 × 10⁻⁴, so according to Eq. (1),

 

Fig. 2. (a) Measured transmission spectrum of DCR-CC filter using swept-laser and measured reflection spectrum of HB-FBG using X-pol. light and Y-pol. light, respectively; (b) Enlarged spectrum of DCR-CC in (a) at ∼1556.3 nm to demonstrate filtering, with fullwidth at half maximum (FWHM) of 8.42 MHz and suppression ratio (SR) of 0.482; SR is defined as the ratio of the height of the first side-peak to that of the main resonant peak, where smaller SR provides better longitudinal-mode filtering.

Download Full Size | PDF

The pure reflection spectrum of the HB-FBG is also shown in Fig. 2(a) in normalized linear scale [18] as measured on a Yokogawa AQ6370D optical spectrum analyzer (OSA). We used linear polarization lights (X-pol. and Y-pol.) for measurement, so the two channels respectively correspond to the two orthogonal polarization principal axes of the PMF (fast axis to X-pol. and slow axis to Y-pol.), with high polarization dependance and high polarization extinction ratio (PER). The HB-FBG reflects two wavelengths naturally separated at two orthogonal linear polarization modes. The two channels both have a ∼0.12 nm (∼14.4 GHz) bandwidth (FWHM) and a central-wavelength spacing of ∼0.272 nm between them (∼1555.822 nm and ∼1556.073 nm, respectively) which is consistent with the calculated value.

As shown in Fig. 2(a), there is one main-channel of the DCR-CC filter located in each channel of the HB-FBG. Considering of the DCR-CC’s filtering FWHM of ∼8 MHz, theoretically, the combination of the HB-FBG and DCR-CC guarantees that the fiber laser performs SLM lasing under a longitudinal-mode spacing of $\Delta \nu$∼4 − 8 MHz corresponding to an allowable main-cavity length of $L$∼ 25.5 − 51.1 m. This is supported by Eq. (2),

Polarization-hole-burning effect (PHB) can arise from the randomly distributed orientations of erbium ions in the glass matrix of an EDF, as well as the selective deexcitation of those ions aroused by a polarized light incident into the EDF [18,45]. Lights with different polarization directions can utilize different subsets of erbium ions to obtain optical amplification, and consequently, the gain competition among different lasing wavelengths can be mitigated. The EDF coiled in the TL-PC can form 1/4, 1/2, and 1/4 wave-plates in turn. A non-negligible bending-induced birefringence, determined by the radii of the EDF and TL-PC’s plate together, is also introduced in the EDF. By controlling the EDF’s polarization axes to parallel to the polarization directions of two incident lights from the two reflection channels of the HB-FBG, the orthogonally spatial distributed erbium ions of the EDF are aroused. This allows for enhanced PHB enabled by the bending-induced birefringence in the EDF, which guarantees the minimum gain competition for two potential lasing wavelengths and stabilizes the dual-wavelength operation of the EDFL.

The fusion-tapered couplers have small polarization dependence which can be magnified in the oscillation sub-rings of the DCR-CC filter. Thus, the DCR-CC is polarization-dependent. A drop-in PC (DI-PC) was placed before the DCR-CC; the gain and loss of the two lasers inside the cavity can be controlled by adjusting the DI-PC. The slight reflectivity difference between the two channels can also introduce polarization-dependent losses for the two lasers at the HB-FBG. By jointly adjusting the TL-PC and DI-PC to change the states of polarization (SOPs) of the oscillating lights inside the laser cavity, switching operations between two single-wavelength and one dual-wavelength lasings can be achieved.

According to our previous work [21], using the superior DCR-CC as the core longitudinal-mode filtering component allows the fiber laser to operate in SLM lasing state with a narrow linewidth of ∼1 kHz. In this work, we developed a new method to more deeply compress the laser linewidth. Based on the femtosecond laser direct writing technique [11,40–42], we inscribed randomly distributed high scattering centers (or scattering points) in the fiber core along the SMF. We used an ytterbium-doped femtosecond fiber laser (YSL Photonics Cor., FemtoYL-20) with a central wavelength of 1030 nm, a pulse width of 382 fs, and a repetition rate of 200 kHz as the direct writing light source. An Olympus UPLXAPO 60× oil-immersion objective lens (numerical aperture: 1.35) focused the laser with a focal spot diameter of <1 µm in the fiber core. The fiber was fixed on an assembled Newport 3D translation stage using two V-grooves. The gap between the objective lens and the fiber was filled with index matching oil to eliminate any distortion induced by the fiber cylindrical surface. Each high scattering center was inscribed through the fiber coating directly with only one laser pulse. To ensure random high scattering distribution, the distance between adjacent scattering centers was set to 5 − 10 cm at random, the average power of femtosecond laser output was chosen randomly from 1.233 W to 1.468 W, and the laser focal spot positions in the fiber core zone were randomly controlled before every inscription.

We fabricated an HSF with maximal length of ∼20 m using the above method. We measured its reflectivity and birefringence distributions using our own polarization-analyzing optical frequency domain reflectometry technique [46], as shown in Fig. 3(a) and 3(b), respectively. As per the measured reflectivity distribution in Fig. 3(a), there are two distinct SMF regions and the HSF region. In the HSF region, randomly distributed high reflection peaks are generally ∼40 dB higher than the reflection background of the SMF with no obvious insertion loss. As shown in Fig. 3(b), there is no obvious birefringence induced by the inscription process in the HSF region compared to the SMF regions, which indicates low polarization dependence in the HSF.

 

Fig. 3. (a) Backscattering distribution and (b) birefringence distribution along the ∼20 m long HSF measured using a polarization-analyzing optical frequency domain reflectometry (PA-OFDR); for comparison, two segments of SMF are fusion spliced respectively at both ends of the HSF for measurements. Photomicrographs of HSF without light (c) and with red laser light (d) incident into the fiber core respectively; in (d) a femtosecond laser induced high scattering center can be clearly seen.

Download Full Size | PDF

We also examined the high scattering center using an Olympus digital imaging microscope based on the UPLXAPO objective, as shown in Figs. 3(c) and 3(d). The scattering center in the fiber core is barely visible in Fig. 3(c), while scattering was clearly observable when a red laser was transmitted through the fiber core, as shown in Fig. 3(d). Based on the laser linewidth compression mechanism of distributed Rayleigh backscattering [32,33]: the continuous scattering process along the fiber synthesizes the statistical characteristics of all random distributed scattering centers, which is equivalent to using a dynamic filtering model with a continuous narrowing linewidth, we believe that our HSF is effective for deeply compressing the linewidth of fiber lasers while a much shorter length is needed compared with the common SMF.

Note that, although the HSF provides the random distributed feedback to compress the laser linewidth, our fiber laser has the essential difference from a random fiber laser, because there is no a standard cavity reflector/mirror to determine the lasing cavity length and longitudinal-mode spacing in a random laser [47] while the HB-FBG is the cavity reflector in our EDFL.

3. Experimental results and discussion

The proposed EDFL system shielded in the isolation box was placed on a common metal optical table. All experiments were conducted under laboratory temperature with the air conditioning constantly running. When the 980 nm pump power was beyond the lasing threshold of the EDFL (∼60 mW), by carefully adjusting the TL-PC and DI-PC, we obtained two single-wavelength operations and one dual-wavelength operation all with high stability. After experimental observation, we identified excellent EDFL output performance when the pump power was controlled in the range of 250-350 mW. We used a pump power of 300 mW for subsequent analysis to demonstrate single- and dual-wavelength operations.

3.1 Single-wavelength operations

The optical spectra of two single-wavelength operations lasing at $\mathrm{\lambda 1}$ and $\mathrm{\lambda 2}$ were measured in turn by the OSA with a resolution of 0.02 nm and a date sampling interval of 0.001 nm, as shown in Figs. 4(a) and 4(b). The 3D spectra were obtained by repeatedly scanning the OSA with a time interval of 5 min over ∼75 min. The two lasers both are stable with pure spectra, and are respectively concentrated at 1555.805 nm and 1556.079 nm, which are consistent with the central wavelengths of two reflection channels of the HB-FBG. The fluctuations of two lasing wavelengths ${f_{\mathrm{\lambda }i}}{\kern 1pt} {\kern 1pt} {\kern 1pt} (i = 1,2)$ and power fluctuations of two lasers ${f_{\textrm{P}i}}{\kern 1pt} ({\kern 1pt} i{\kern 1pt} = 1,2)$ are respectively less than 0.007 nm and 0.473 dB, as shown in the legends of Figs. 4(a) and 4(b). The OSNRs of the two lasers both exceed 85 dB, which is state-of-the-art level for SLM or single-frequency fiber lasers. The two lasers also have orthogonal polarization states due to the polarization-dependent two-channel HB-FBG. We did not observe any side mode peak except the lasing wavelength peak in every spectrum, so we conclude that the two lasers have high PER between each other.

 

Fig. 4. Spectra of single-wavelength switchable operations lased at (a) λ1 and λ2, as-measured by OSA over ∼75 min. ${f_{\mathrm{\lambda }i}}{\kern 1pt} {\kern 1pt} {\kern 1pt} (i = 1,2)$: fluctuation of lasing wavelength at $\mathrm{\lambda }i$; ${f_{\textrm{P}i}}{\kern 1pt} ({\kern 1pt} i{\kern 1pt} = 1,2)$: fluctuation of power lasing at $\mathrm{\lambda }i$. OSNR: optical signal to noise ratio.

In each image, 15 repeated spectra are shown measured with an interval of ∼5 min.

Download Full Size | PDF

We measured the linewidths of two lasers using the delayed self-heterodyne method. We constructed the measurement system using a fiber Mach-Zehnder interferometer (MZI, with a 200 MHz acoustic optical modulator (AOM) and a ∼100 km SMF in two arms respectively), a 400 MHz photodetector (PD, Thorlabs PDB470C), and a radio frequency (RF) electrical spectrum analyzer (ESA, Keysight N9010A) [18]. The delayed self-heterodyne beating spectra for λ1 lasing and λ2 lasing of the EDFL with the 20 m HSF were measured in 199.95-200.05 MHz as given with blue lines in Fig. 5(a) and 5(c). For better curve-fitting, enlarged plots were drawn as shown in Fig. 5(b) and 5(d) under a frequency range of 199.98-200.02 MHz. Lorentz fitting was performed for two data lines with high Adj. R-Square parameters of 0.9919 and 0.9924, respectively. The broadening effect of the 1/f frequency noise from the long delay fiber is most pronounced near the center of the self-heterodyne lineshape. A more accurate estimate of Lorentz part of the linewidth (intrinsic linewidth) of the laser could be obtained from the bandwidth 20 dB down from the maximum [48]. We calculated the linewidths for the two lasers as 1/20 times the 20 dB bandwidths of the fitting curves as 125.1 Hz and 149.9 Hz, respectively. Compared to the same-level linewidth obtained using the SMF as the backscattering medium [33,36], the laser cavity length used much less HSF at only 20 m.

 

Fig. 5. (a)/(c) Delayed self-heterodyne beating spectrum variation with respect to length of the HSF and (b)/(d) Lorentz fitting of spectrum measured for λ1/ λ2 lasing, under using the 20 m HSF; Measured linewidth as a function of HSF length for λ1 lasing (e) and λ2 lasing (f).

The side peaks marked by the green circles are induced by the partial coherence mixing of MZI’s two arms (the frequency spacing between adjacent side peaks is close to MZI’s FSR).

Download Full Size | PDF

To validate the linewidth compression effectiveness of the HSF, we gradually cut down its length and continually measured the linewidths of two lasers. Figures 5(a) and 5(c) show changes in the delayed self-heterodyne beating spectra with the residual length of HSF for λ1 lasing and λ2 lasing. For each laser, the spectrum width broadens as the HSF length decreases. The detailed relationships between linewidth and HSF length for the two lasers are also given in Figs. 5(e) and 5(f). The linewidths narrow significantly as the HSF length increases. For λ1 lasing and λ2 lasing, respectively, the linewidths are deeply compressed from 1120.4 Hz to 125.1 Hz and from 1084.8 Hz to 149.9 Hz. In theory, our system with the ∼100 km long delay fiber has a resolution of ∼2.1 kHz for measuring the self-heterodyne beating spectrum [49]. The measurement limit of minimal laser linewidth should be ∼105 Hz, theoretically. Although using a longer delay fiber can improve the resolution, the 1/f frequency noise from the ultra-long delay can seriously influence the linewidth measurement accuracy. In addition, a narrower linewidth can induce more serious partial coherence mixing of the MZI’s two arms, which can render the estimation intrinsic linewidth of the laser inaccurate from the 20 dB bandwidth of the fitting curve. Therefore, we did not investigate the linewidth compression limit of our fiber laser with an HSF longer than 20 m.

The power spectral density (PSD) of frequency noise for each laser was measured with a phase noise analyzer (PNA, Rohde & Schwarz FSWP26) based on a fiber MZI with the 200 MHz AOM and a 100 m long SMF in two arms respectively. The laser output passes through the MZI to transform the laser’s frequency noise to the MZI’s differential phase noise, which is modulated on a 200 MHz carrier frequency. The MZI’s output was detected by the 400 MHz PD, the output of which was then monitored by the PNA over a range from 10 Hz to 1 MHz. Finally, according to the relationship between frequency noise and phase noise (Eq. (3)), we determined the PSDs of frequency noise for λ1 lasing and λ2 lasing. In Eq. (3), ${S_\phi }(f )$ (rad²/Hz) and ${S_\nu }(f )$ (Hz²/Hz) denote the PSDs of phase noise and frequency noise in linear scale, respectively, and $\Delta L$=100 m denotes the length difference between two arms of the MZI.

The frequency noise spectra of λ1 lasing and λ2 lasing with and without the 20 m HSF were measured as shown in Figs. 6(a) and 6(b). For comparison, in each figure the frequency noise spectra of a single-frequency fiber laser (NKT Photonics, E15), a DFB diode laser module (Han’s laser, RP-MP-10-0100-02), and a tunable laser source (Yenista, T100S-HP) with claimed linewidths of 100 Hz, 100 kHz, and 400 kHz are also shown. Our fiber laser’s frequency noise performance is superior to other lasers regardless of the lasing wavelength. Because our measurement system does not include any specialized temperature control shield or particular vibration and acoustics insulation protections, there is substantial noise at Fourier frequencies less than 1 kHz measured for our EDFL; we attribute this mainly to acoustic noise from the measurement system itself and its surroundings.

 

Fig. 6. Frequency noise spectra measured with correlation delay self-heterodyne method for (a) λ1 lasing and (b) λ2 lasing under with 20 m HSF and without HSF, respectively; for comparison, the frequency noise spectra of a NKT E15 single-frequency (SF) fiber laser, a commercial DFB diode laser module and a Yenista T100S-HP laser source measured using the same method are plotted in each figure. ${S_{0,\mathrm{\lambda }i}}({i = 1,2} )$ and ${S_{0,\mathrm{\lambda }i\textrm{ - HSF}}}({i = 1,2} )$ denote the white frequency noises of corresponding lasers.

Download Full Size | PDF

At very high Fourier frequencies, the frequency noise of the lasers is mainly the white quantum-noise floor, which reflects the lasers’ intrinsic linewidths as per $\pi {S_{0,\mathrm{\lambda }i\textrm{ - HSF}}}({i = 1,2} )$ [50]; ${S_{0,\mathrm{\lambda }1\textrm{ - HSF}}}$ and ${S_{0,\mathrm{\lambda 2\ -\ HSF}}}$ denote the white noise floors for λ1 lasing and λ2 lasing, respectively, with HSF in the laser cavity. As marked in Figs. 6(a) and 6(b), the ${S_{0,\mathrm{\lambda }1\textrm{ - HSF}}}$ and ${S_{0,\mathrm{\lambda 2\ -\ HSF}}}$ are 0.8 Hz²/Hz and 2.7 Hz²/Hz, respectively, indicating intrinsic linewidths of 2.5 Hz and 8.5 Hz. By contrast, when the 20 m HSF was removed, the laser frequency noise at low frequencies did not change significantly; this further suggests that noise mainly originated in the measurement system and its surrounding environment. At high frequencies, we observed a marked increase in the white noise floor. The white noise floors ${S_{0,\mathrm{\lambda }1}}$ and ${S_{0,\mathrm{\lambda 2}}}$ are 4.5 Hz²/Hz and 12.3 Hz²/Hz, respectively, indicating that the intrinsic linewidths for λ1 lasing and λ2 lasing are 14.1 Hz and 38.6 Hz. Therefore, the HSF effectively compresses the intrinsic linewidths of the fiber lasers. In addition, the results also prove that our DCR-CC filter itself was capable of achieving an ultra-narrow intrinsic linewidth for a fiber laser.

The SLM lasing stability of the proposed EDFL with the HSF was investigated on a scanning Fabry-Pérot interferometer (Thorlabs, SA200-12B) with an FSR of 1.5 GHz and a resolution of 7.5 MHz, as shown in Figs. 7(a) and 7(b) for λ1 lasing and λ2 lasing, respectively. We find that the EDFL functioned in stable SLM states for both single-wavelength operations without mode-hopping unless a serious disturbance was imposed on the shield box. The mode-hop characteristics were also monitored by the delayed self-heterodyne method with a 100 km long delay fiber in the MZI. As shown in Fig. 7(c), using the maximum-hold mode of the ESA in a measurement time range of ∼30 min, there is only one strong beating signal at 200 MHz induced by the AOM. The use of 100 km SMF provided enough time delay to capture any mode-hopping, so the results in Fig. 7(c) indicate that the proposed EDFL is capable of working in SLM operation without mode-hopping at both lasing wavelengths over a long period of time. The SLM selection capability of the DCR-CC filter was also confirmed by replacing it using an identical effective length of SMF fiber in the laser cavity; we measured the radio frequency spectra of both lasers based on the self-homodyne method using the 400 MHz PD and ESA over ∼10 min, as shown in Fig. 7(d). Numerous multi-longitudinal-mode beating signals were captured with a frequency spacing of 3.89 MHz.

 

Fig. 7. Longitudinal-mode characteristics, measured by a scanning Fabry-Pérot interferometer, for λ1 lasing (a) and λ2 lasing (b), respectively; RF spectra measured by ESA with maximum-hold mode for both lasers, (c) using delayed self-heterodyne method in ∼30 min in a range of 0-250 MHz and (d) using self-homodyne method in ∼10 min in a range of 0-100 MHz when DCR-CC filter replaced by SMF with a length identical to its equivalent cavity length.

Download Full Size | PDF

The RIN spectra of two single-wavelength operations were measured using a 400 MHz PD and the ESA. An oscilloscope (Tektronix, TDS2024C) was also used to measure the direct voltage of the PD output. The laser output power was attenuated to 500 µW to prevent saturation of the PD. The laser outputs with and without the 20 m HSF in the laser cavity were measured for λ1 lasing and λ2 lasing as shown in Figs. 8(a) and 8(b) over a frequency range of 0-5 MHz. The shot noise limit (SNL) for each laser was also calculated [18]. Inset 1 and inset 2 show data in the range of 0-200 kHz which includes relaxation oscillation peaks. The measured RIN is very close to the SNL at frequencies >3 MHz in each figure, while the insertion of the HSF in the laser cavity does not influence the RIN spectrum despite a significant change in laser cavity length. The RINs at frequencies >3 MHz for λ1 lasing and λ2 lasing are both less than −150 dB/Hz. HSF insertion shifts the relaxation oscillation peaks of the two lasers from ∼57 kHz to ∼32 kHz by increasing the laser cavity length.

 

Fig. 8. RIN spectra of single-wavelength operations with (blue solid-line) and without (red dashed-line) 20 m HSF lased at (a) λ1 and (b) λ2, in 0-5 MHz using RBW of 10 kHz for ESA with corresponding SNL; Insets show same measurements in 0-200 kHz using RBW of 100 Hz with relaxation oscillation peaks.

Download Full Size | PDF

To validate the SLM selection capability of the DCR-CC filter and HSF in the laser cavity, we inserted an SMF with gradually increasing length between the HSF and HB-FBG. We measured the self-homodyne beating spectrum of λ1 lasing output after every length change of SMF. When the inserted SMF length is 16 m, the SLM cannot be maintained (Fig. 9). An obvious beating signal at frequency of 2.44 MHz was captured corresponding to a laser cavity length of ∼84 m. The HSF lengthens the allowable cavity length for SLM operation obtained according to the passband of the DCR-CC filter (Section 2), but the SLM cannot be guaranteed beyond a certain laser cavity length.

 

Fig. 9. Self-homodyne RF spectrum of λ1 lasing measured using maximum-hold mode of ESA in 5 min with 16 m SMF in laser cavity between HB-FBG and HSF; EDFL cannot maintain stable SLM operation.

The beating signal at frequency 2.44 MHz corresponds to a laser cavity length of ∼84 m.

Download Full Size | PDF

3.2 Dual-wavelength operation

Dual-wavelength operation is possible by simply adjusting the two PCs, as a benefit of the enhanced PHB effect established in the EDF (Section 2). The extremely stable optical spectra, measured by repeatedly scanning the OSA in a time span of ∼50 min, are shown in Fig. 10(a). The wavelength and power fluctuations are <0.003 nm and <0.421 dB. The OSNR is as high as >83 dB, which to the best of our knowledge is unprecedented. We verified the stable SLM lasing for both lasers on a scanning F-P interferometer, as shown in Fig. 10(b). Over a monitoring span of ∼10 min in a quiet test environment, we did not observe any mode-hopping unless the shield box was intentionally manipulated.

 

Fig. 10. Spectra of dual-wavelength operation measured by OSA over ∼50 min; (b) Longitudinal-mode characteristic measured by scanning Fabry-Pérot interferometer; Delayed self-heterodyne beating spectra of dual-wavelength lasing with (c) and without (d) 20 m HSF in the laser cavity.

Each figure ((c) and (d)) shows Lorentz lineshape fitted to measured data.

Download Full Size | PDF

The laser output linewidth in dual-wavelength operation was also measured by the delayed self-heterodyne method before and after removing the 20 m HSF as shown in Figs. 10(c) and 10(d). The laser linewidth increased from 144.2 Hz to 1228.5 Hz after the HSF was removed, indicating that for simultaneous dual-wavelength lasing, the HSF still effectively compressed the linewidth of each laser.

4. Conclusion

In this work, we developed a novel laser linewidth compression mechanism for fiber lasers based on a femtosecond laser-induced HSF. Random artificial scattering centers were inscribed in the SMF with an enhanced scattering intensity level ∼40 dB higher than the SMF’s natural Rayleigh backscattering background. Using a low-cost SLM narrow linewidth dual-wavelength EDFL for demonstration, we determined that the laser output linewidth can be reduced with increasing HSF length. Using the delayed self-heterodyne method with a 100 km fiber delay-line for measurements, we find the linewidths of two single-wavelength lasing states of the EDFL can be compressed from 1120.4 Hz to 125.1 Hz and from 1084.8 Hz to 149.9 Hz, respectively, with a 20 m HSF. Our EDFL shows better frequency noise performance than three commercial single-frequency lasers. The PSDs of frequency noise further show that the intrinsic linewidths of two lasers can be compressed from 14.1 Hz to 2.5 Hz and from 38.6 Hz to 8.5 Hz, respectively, with the HSF. We also find that the HSF is equally effective for linewidth compression (from 1228.5 Hz to 144.2 Hz) in case of simultaneous dual-wavelength lasing output.

Beyond the superior linewidth features, our dual-wavelength switchable SLM EDFL shows outstanding performance including stability in the optical spectrum and mode-hop-free SLM lasing. The RIN is as low as <−150 dB/Hz, approaching the SNL, at frequencies >3 MHz. The EDFL can output lasers with the state-of-the-art OSNRs in two single-wavelength operations (>85 dB) as well as in dual-wavelength operation (>83 dB), unlike similar SLM fiber lasers reported previously. The proposed linewidth compression method for fiber lasers and the high-performance dual-wavelength switchable SLM ultra-narrow linewidth EDFL are suitable for practical application. In the future, we will optimize the HSF fabrication process for obtaining the best possible laser linewidth compression effect for SLM fiber lasers in different wavelength bands.